NV GTX 460 1GB SLI vs. ATI HD 5870 CFX Redux @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,601
NV GTX 460 1GB SLI vs. ATI HD 5870 CFX Redux - We re-visit GeForce GTX 460 1GB SLI vs. Radeon HD 5870 CrossFireX performance using new Catalyst drivers and new CrossFireX Application Profile 10.8a. We also have a new BETA ForceWare driver in play and we will now see how these video card configurations compete with updated software.
 
Thank you for this latest series of write ups, I'm looking into potentially buying these cards right now and it's great to know what all the options are.
 
Good job for AMD and terry to improve performance. Shame it took so many driver updates and Hardocp to point it out to them. Either way I still say the drivers aren't optimized enough as a 5870 crossfire setup should trade blows with GTX 480 SLi and unfortunately it doesn't. AMD should be able to squeeze out much more performance and their customers should be pissed. Paying premium for 2 5870s to have it trade blows with 2 GTX 460 1gbs. I know AMD wins this battle now but on how many games is a good question. We know in avp and bfbc2 it does, possibly a few others. It's just not impressive enough for me. I know the hardware is there.

Shouldn't be this way, AMD has a fantastic employee in Terry, however they need like 20 more clones of him.
 
How about an [H] comprision of 3x 460 1GB vs 2x 5870 since the price for each setup would be the same.
 
Well since you can't Tri-SLI 460's together, I think that would be rather difficult for them to test, no?
 
lol good thing the only game i am playing right now is metro and i ditched my 5850s for 460s.
 
How do the forceware 260.52 line up to the beta 260.63? if its a major improvement it would be nice to see a update using those
 
thanks [H] for redoing all these reviews , now my 5870 gain value back again :)
 
While I appreciate the redux alot......one thing really rubs me the wrong way.

Several weeks ago ATI sucked in the games you demonstrated today, specifically BC2 and AVP and we all know that Metro has pretty much always shone on nvidia products.

Now ATI specifically polishes drivers for BC2 and AVP and surprise.....they're faster, and there is essentially no change in Metro.

Why not pick three random games and use the same driver profiles.....see what ATI does without a ringer.

I fully understand the comparison, but it's sort of like going to ATI and asking them to focus on just these games and not doing the same for nvidia.

Yup.....I have EyeFinity and I have 5870 E6 and I also have Surround with GTX 260, so I am not anybody's fanboy.....I just dont think this is a fair fight.
I appreciate the work you do, I appreciate that ATI sucked it up and fixed THOSE TWO GAMES........but what else do they have?
I quit playing both of those this Spring.
 
magoo said:
While I appreciate the redux alot......one thing really rubs me the wrong way.

Several weeks ago ATI sucked in the games you demonstrated today, specifically BC2 and AVP and we all know that Metro has pretty much always shone on nvidia products.

Now ATI specifically polishes drivers for BC2 and AVP and surprise.....they're faster, and there is essentially no change in Metro.

Why not pick three random games and use the same driver profiles.....see what ATI does without a ringer.

I fully understand the comparison, but it's sort of like going to ATI and asking them to focus on just these games and not doing the same for nvidia.

Yup.....I have EyeFinity and I have 5870 E6 and I also have Surround with GTX 260, so I am not anybody's fanboy.....I just dont think this is a fair fight.
I appreciate the work you do, I appreciate that ATI sucked it up and fixed THOSE TWO GAMES........but what else do they have?
I quit playing both of those this Spring.
Click to expand...

Good post, I was thinking similar things. I did see that they supposedly fixed 2nd gpu not working in crossfire also they supposedly fixed starcraft 2 crossfire and AA. I agree that by checking only those 3 games AMD was able to put a bandaid on an infection i.e. it looks good on the outside as a solution but what is happening on the inside. If we pull the curtain will we find that their crossfire is borked on over 30 games and since this review they focused and fixed 2 of the 30 games.

2nd guessing will probably be the norm for amd until they restore faith in their drivers.
 
So, months after these games have come out, and I've already played through them and couldn't give a crap about them anymore, ATI finally has a driver that beats budget nvidia cards in SLI on a couple of hand picked games with their top end card in crossfire. Gee...for some reason, I'm not impressed. I'm done with ATI as far as multi-card setups go. Crossfire is a joke compared to SLI, and these tests just further illustrate that. I've been running my 4870's in crossfire for two years now, and I've never been real happy with the performance. In fact, there have been several games that actually performed better when I disabled crossfire. Not what I want to deal with when spending hundreds of dollars to run multiple video cards.
 
magoo said:
While I appreciate the redux alot......one thing really rubs me the wrong way.

Several weeks ago ATI sucked in the games you demonstrated today, specifically BC2 and AVP and we all know that Metro has pretty much always shone on nvidia products.

Now ATI specifically polishes drivers for BC2 and AVP and surprise.....they're faster, and there is essentially no change in Metro.

Why not pick three random games and use the same driver profiles.....see what ATI does without a ringer.

I fully understand the comparison, but it's sort of like going to ATI and asking them to focus on just these games and not doing the same for nvidia.

Yup.....I have EyeFinity and I have 5870 E6 and I also have Surround with GTX 260, so I am not anybody's fanboy.....I just dont think this is a fair fight.
I appreciate the work you do, I appreciate that ATI sucked it up and fixed THOSE TWO GAMES........but what else do they have?
I quit playing both of those this Spring.
Click to expand...

What games would you like us to test now. We have another followup planned, but are sort of stuck on the games. Mafia II....then what that really pushes the cards?
 
Darksword said:
How about an [H] comprision of 3x 460 1GB vs 2x 5870 since the price for each setup would be the same.
Click to expand...

How? 3x460 is not possible. Also it doesn't make sense to compare Tr-SLI/CF setups for reviews because most people will not have that kind of setup. Reviewers usually bench setups that are the most common.
 
Kyle- any chance you could bench Starcraft 2? I know that its also one thats had Crossfire issues in the past.
 
It really rubs me the wrong way that AMD/ATI only updates it's drivers to fix scaling issues if they are brought to the forefront by a review site. They should be looking at games and saying "Hey, this game has really shitty scaling on xfire, how about we go fix that" not waiting till someone points it out. It's like that issue a while back where the laptop had 2 graphics cards in crossfire, but didn't enable crossfire because they never updated the drivers.



Kyle-
You were asking for games to run these cards on to push them, I'd love to see an SC2 and FFXIV. Although there's a good bit of speculation that those both may be somewhat CPU limited. I know I saw one review on SC2 that was showing the FPS increases by both increasing CPU clock and from going from dual to tripple channel memory. I'd be very interested on seeing some attention paid to it.
 
In the conclusion you state that the 5870 cfx is faster than the 460 SLI setup in BC2. Are you referring to the data numbers or are you referring to "actual gameplay"? Since the minimum is the same in the results, I was curious if you could see an actual difference when playing both setups?

Thanks
 
Interesting article again. Would it be possible for another 480/SLI review and how it compares again vs 5870 vs 470 vs 460?
 
When I saw the original article and saw the 460's lay the 58xx's down so hard, it really made me irritated at the ATI team. Seeing the huge reversal in performance is awesome though, and really makes it clear how important good drivers are. I hope the AMD guys have solid drivers ready for the 6000 series launch. Awesome hardware, but things like this shake my faith in AMD/ATI graphics.
 
glad to see ati finally get the driver updates rolled out. it just brings the competition closer between nvidia and ati. win win for consumers.
 
I'd like to see an article that inspects weather or not Metro2033 is actually of any use as a benchmark when putting Nvidia against AMD. Happy to see profiles coming out and doing a real good job but Metro2033 still seems to be doing a Crysis of a job on the brand vs brand front.
 
WorldExclusive said:
How? 3x460 is not possible. Also it doesn't make sense to compare Tr-SLI/CF setups for reviews because most people will not have that kind of setup. Reviewers usually bench setups that are the most common.
Click to expand...

Well if that was true, [H] would only focus on benchmarks with GMA chipsets :D:D
 
Well if I were recommending for a friend who wanted to save money, I'd tell him to buy 460 SLI every time. The performance difference does not justify the nearly $300 price differece (after rebates) of a 5870CF vs a 460SLI setup. $700 vs $400.
 
3DChipset said:
In the conclusion you state that the 5870 cfx is faster than the 460 SLI setup in BC2. Are you referring to the data numbers or are you referring to "actual gameplay"? Since the minimum is the same in the results, I was curious if you could see an actual difference when playing both setups?

Thanks
Click to expand...

minimum fps are the same doesnt mean the smoothness feel the same or different, it doesnt really mean anything until the fps drop and rise like crazy in graph.

The minimum fps usually occur when there is a loading or check point, the fps drops a sec then comes back for all setup.
 
I think it's funny that people get so crazy about comparing these cards. the 5870 is a YEAR old now , wow it is close in performance to a brand new card. not a big deal IMO. I mean AMD will have new cards out in a short time here that will jump the level up once again. just buy what works best for you and enjoy your games. on a side note AMD does need to get there shit together with the drivers lately. really tired of HDMI scaling issues.
 
Last edited:
It seems my decision to sell ONE of my 5870's for $350 was a smart move. I spent an additional $90 out of pocket and got me a pair of Asus ENGTX460 1gig for SLI.

I'm able to run both a 850/2200 without issue.

When the new 6870 comes out, I will dump the the 460's and get me the new ATI card. Won't miss a beat and should be able to upgrade right away. A hassle yes, but worth it.
 
I feel sorry for those who bought the "SLI scales better" BS because one accidental driver from AMD.
 
vjcsmoke said:
Well if I were recommending for a friend who wanted to save money, I'd tell him to buy 460 SLI every time. The performance difference does not justify the nearly $300 price differece (after rebates) of a 5870CF vs a 460SLI setup. $700 vs $400.
Click to expand...

Do know that both 5850 and 5850 SLI are both faster 460 and 460 SLI, the difference is about $30 for a single card and $60 for SLI.

The 460 is poised in between 5830 and 5850, nothing more.
 
I don't think you are taking into account that everyone OTHER than review web sites are running these cards at 800 - 900mhz / 2200 - 2300.

I've had both 5870 crossfired and now 460's SLI'd and my 460's are at the same level / slightly faster than crossfire 5870's.
 
1) HardOCP sees the 5870's CF as pathetic enough that it trades blows with cards half its price in the mid-range by benching specific new titles.
2) AMD releases a new driver, tailored to improve performance specifically in those titles, then says, "Hai, HardOCP, everything's gud now, our drivers are gud!"
3) HardOCP checks it, confirms that AMD did in fact improve (aka "fix") performance in those specific, often-used-as-benchmarks games and declares AMD great again.

1) nVidia optimizes for certain games and benchmarks.
2) HardOCP benchmarks other games and calls nVidia on it by renaming the executables to trick certain games based on certain engines to turn on or off optimizations.
3) HardOCP cries foul and declares nVidia the evil to end all evils, that optimizations that "fix" performance on specific titles and/or benchmarks alone are evil.

Oh, how times have changed. Back in the day, it was evil. Today, it's just a company updating its drivers. Come on. They update a few specific titles with optimizations and you take their hand and let them lead you down the golden path to see the paradise they've laid out for you.

After they said they went to the trouble of "fixing" said games, you go and use other games. You don't go back and make an article showing how awesome AMD is at optimizing for specific games when necessary.

Plus, what about the next big hit? Is that also going to require waiting months after the game is even relevant to most gamers before it performs appropriately? Isn't it odd that nVidia can come up with a driver that's tip-top in performance while AMD's still wringing their hands on games after months of release? Where's the outrage, HardOCP? Where's the crying foul at specifically "fixing" games that you use as benchmark while doing nothing for what will come in the future?
 
As always, another great article. The fact that there's even a comparison of nVIDIA cards that cost 1/3 less ($480 vs $710) less than ATI cards says a lot about the price/performance of the GTX 460.
 
Makes me glad I sold my 5870 for $350 then picked up two 470's for $250 a piece. This was also before the cheap dp -> dvi adapters so it made my eyefinity/surround setup possible for $100 less, so my net out of pocket expense was $50 if I wanted to run 3 screens.
 
HisDivineShadow said exactly what I was thinking. Something stinks about this driver "optimization" and sounds a lot like what Nvidia and ATI have done in the past to help get their cards sold. If they truly are optimizing and fixing broken code, great, but if this is all a white wash over the brown crap, then they dont deserve any praise at all.

Personally it is probably good enough for the average joe to see that their game runs better, but I am just wondering what the cost was. There is no free lunch yadda yadda, what did they do to get their improved numbers. We may never know.
 
Any chance you can consider framerate irregularity (aka "microstutter") when comparing the two setups? This phenomena has a real impact on the performance of a setup, which does not show up in raw-FPS benchmarks.

If you're interested I wrote a program to quantify the apparent performance reduction due to microstutter, from FRAPS benchmarks (program and related thread), although you could easily use your own algorithm if you like.

Microstutter really can make a game running at 60fps seem like 40fps, in the worst cases, and it's a subtle effect that you can't really spot straight off (in the same way as, say, tearing or paging/hitching). I'd love to see review sites taking this seriously, but most are just interested in simple average FPS comparisons between setups. These don't tell the whole story, especially when it comes to multi-GPU setups.
 
Arseface said:
Any chance you can consider framerate irregularity (aka "microstutter") when comparing the two setups? This phenomena has a real impact on the performance of a setup, which does not show up in raw-FPS benchmarks.

If you're interested I wrote a program to quantify the apparent performance reduction due to microstutter, from FRAPS benchmarks (program and related thread), although you could easily use your own algorithm if you like.

Microstutter really can make a game running at 60fps seem like 40fps, in the worst cases, and it's a subtle effect that you can't really spot straight off (in the same way as, say, tearing or paging/hitching). I'd love to see review sites taking this seriously, but most are just interested in simple average FPS comparisons between setups. These don't tell the whole story, especially when it comes to multi-GPU setups.
Click to expand...

Good point, microstutter can be a pain the ass! :mad:
 
Arseface said:
Any chance you can consider framerate irregularity (aka "microstutter") when comparing the two setups? This phenomena has a real impact on the performance of a setup, which does not show up in raw-FPS benchmarks.

If you're interested I wrote a program to quantify the apparent performance reduction due to microstutter, from FRAPS benchmarks (program and related thread), although you could easily use your own algorithm if you like.

Microstutter really can make a game running at 60fps seem like 40fps, in the worst cases, and it's a subtle effect that you can't really spot straight off (in the same way as, say, tearing or paging/hitching). I'd love to see review sites taking this seriously, but most are just interested in simple average FPS comparisons between setups. These don't tell the whole story, especially when it comes to multi-GPU setups.
Click to expand...

Problem with trying to measure microstutter, is that it is not always reproducible. Two people can have very similar or even identical rigs, and one suffers microstutter and the other does not. Different people also have a different threshold of where microstutter becomes noticeable.
Microstutter exists and is a problem, and can be measured, but just not in a way that can be very meaningful as it relates to [H]'s readers imho. I personally, have never noticed microstutter. Not with Voodoo2 sli, 6800 sli, 7800 sli, 1950 xfire, 8800, sli, or 5870 xfire.
 
GORANKAR said:
Problem with trying to measure microstutter, is that it is not always reproducible. Two people can have very similar or even identical rigs, and one suffers microstutter and the other does not. Different people also have a different threshold of where microstutter becomes noticeable.
Microstutter exists and is a problem, and can be measured, but just not in a way that can be very meaningful as it relates to [H]'s readers imho. I personally, have never noticed microstutter. Not with Voodoo2 sli, 6800 sli, 7800 sli, 1950 xfire, 8800, sli, or 5870 xfire.
Click to expand...

Read the thread I have on the subject. Like I said, it isn't a "noticeable" phenomenon, in that you look at a game scene and say "hey, this is stuttering". It just works to reduce the apparent framerate down from what the thread counter says. In other words, 60fps with microstutter is not the same performance as 60fps with regular output.

As for reproducibility; I have tested the program on a wide range of setups, and microstutter is always present to some degree when multi-GPU setups are rendering in AFR mode. Sure it varies from game to game, and across different configurations (just like framerate or anything else), but it is certainly reproducible, and while microstutter scores may vary by a few % with similar rigs, in identical circumstances it is always observed to the same magnitude.

The only circumstances in which microstutter disappears are:

1. CPU-limitation (here the output of the GPUs syncs to that of the CPU, as they are waiting for the CPu to finish its workload. Clearly CPU output is regular)
2. When vsync is enabled.

Again, like I said, microstutter isn't something you "just notice", but it is still present, and is eating away at your performance. If you run my program on some FRAPS benchmarks you will see by how much.
 
You must log in or register to reply here.
Back
Top